Neurocognitive impairment is believed to affect alcoholism treatment process and outcome. However, alcohol researchers have had difficulty empirically validating this relationship between impairment and outcome, often finding weak and inconsistent support. This competitive renewal application seeks to extend a Rutgers CAS research program studying neurocognitive impairment in persons treated for alcohol use disorders. The unmet need to examine neurocognitive moderation models has often been highlighted by alcohol researchers and is strongly supported by the literature on the relation of neuropsychological impairment to behavioral outcomes following treatment for brain injuries of multiple etiologies. We have found support for a hypothesized model wherein neurocognitive impairment moderated the relation of robust change processes to intensity of alcohol and other drug use 6 months after treatment. Self-efficacy, commitment to abstinence, and AA affiliation were strong predictors of outcome in unimpaired persons, but only weak predictors of outcome in those who were impaired. Results supported a threshold model of impairment which suggests that only severe neurocognitive impairments may affect moderation. The goal of this competitive renewal application is to develop heuristic models of brain-behavior relations that adequately characterize the effect of alcohol-related neurocognitive impairment on treatment outcome through secondary analysis of three existing longitudinal data bases which overlap substantively in constructs and measures. We will use two convergent analytic methods to contrast direct effect, mediation, and moderation models of neurocognitive influences on alcohol- and drug-specific, and psychosocial outcomes within a conceptual framework informed by the traumatic brain injury rehabilitation literature. By examining the generality of results across different addictions treatment populations, we can provide a more robust test of putative neurocognitive models of treatment outcome than otherwise would be possible without costly collection of new longitudinal data. Results will advance understanding of mechanisms which support good outcomes in neurocognitively impaired individuals and will have theoretical implications for cognitive models of relapse.